Developing and applying innovative methods to integrate high-dimensional multi-omics data with clinical outcomes data
Statistical methods that assess multiple disease factors offer insights that advance research and clinical treatment. To aid in this advancement of discovery and care, our laboratory develops and applies innovative statistical methods to evaluate multiple forms of molecular data. Through our work, our goal is to determine what patient- and disease-specific factors drive disease progression and outcomes in pediatric catastrophic disease.
The ability to simultaneously assess and establish associations between multiple forms of high-dimensional molecular data serves many benefits in the realms of research and clinical care. Through our methodology development and application, we work to determine what molecular characteristics drive prognosis, and we continue to translate these methodology advancements into better clinical care through our rich collaborations with biologists and clinicians.
Our work in methodology development focuses on three areas that offer novel approaches to help us understand how we can best establish associations between complex sets of multi-omics data.
Bootstrap evaluation of association matrices (BEAM)
The BEAM method is used to construct a matrix that establishes associations between different variables within large datasets. Once we amass a computational cloud of data points in space, we can measure if there is a meaningful association between all the features in relation to patient survival. We eventually hope to link this method to what we know about gene and drug interactions available via public databases so we can begin to make therapy suggestions.
Genomic random intervals (GRIN)
We use the GRIN method to find genes that have an overabundance of genomic abnormalities in tumor cells. Because this method examines all types of genomic lesions simultaneously, we can discover important, but cryptic, genomic abnormalities that manifest in different ways. Our goal is to tie GRIN to gene expressions and clinical outcomes.
Distance-based methodologies operate by computing distance matrices between pairs of data points. Since distance matrices illustrate how different the profiles between two data points are, we use this information to establish distance correlation statistics between these two points. Distance correlation statistics help determine whether there is a meaningful association, in terms of expression of clinical outcomes, in these distance correlations. The power of this method lies in its ability to detect an entire class of associations among variables that current methodologies based on linear calculations cannot detect.
The development and application of all these methodologies broadens our ability to discover meaningful associations that were previously invisible in large datasets.
Through our clinical collaborations, we can apply our expertise to employ existing and novel methodologies that inform clinical care and research. Most of our clinical collaborations focus on pediatric leukemias, especially acute myeloid leukemia (AML). We have provided biostatistical services for the last three frontline AML clinical trials and several relapse trials.
In the realm of AML pharmacogenomics, our interest lies in defining genomic features that impact outcomes through pharmacologic mechanisms. We developed a pediatric leukemic stem cell score—which are currently validating in many cohorts from around the world—that quantifies the extent of stem cells within pediatric AML. Because a high stem cell score associates with poor outcomes, our work offers insights that can help guide appropriate treatment.
To guide clinical insight in acute lymphoblastic leukemia (ALL) cases, we apply our genomic random intervals (GRIN) method. We have found several genes that have an overabundance of genomic lesions that drive the development and progression of this disease.
Beyond our active clinical collaborations focused on pediatric leukemias, we also provide biostatistical consultation services for investigators studying brain and solid tumors.
Our extensive involvement in clinical trials and omics methodology development defines the character of our laboratory, and we take pride in our ability to make our omics methods research clinically relevant in the treatment of pediatric catastrophic disease.
Dr. Stanley Pounds is a biostatistician with extensive collaborative experience in laboratory and clinical pediatric oncology research. A Member of the St. Jude Faculty, he received his PhD in Statistics from Texas A&M University and serves as Director of Biostatistics Courses at the St. Jude Graduate School of Biomedical Sciences. He is currently developing a biomedical data science program within the St. Jude Graduate School. As a recognized expert in the development and application of innovative data analysis methods for applications in cancer genomics and pharmacogenomics, Pounds leads his laboratory in statistical research that makes practical contributions toward enhancing rigor and impact of collaborative biomedical research.
A team of biostatisticians who make omics methods research clinically relevant through involvement in both clinical trials and omics methodology development.